Color constancy refers to the stability of object color appearance despite variations in the ambient lighting and surfaces present in an image. We propose to test a series of hypotheses about how human observers obtain a stable judgment of the color appearance of objects. In the initial period we plan to use multi-colored images presented on a CRT. The controlling software is designed to permit us to specify the spectral power distribution of the ambient lighting, and surface reflectance functions at different locations in the image. We will measure the achromatic locus (a curve in color space along which the display appears white) as a funciton of (1) the spatial position in the display, (2) the spectral power distribution of the ambient light used to generate the display, and (3) the selection of surface reflectance functions in the display. We play to use measurements of the achromatic locus to test two classes of theories of spatial integration of color information. Buchsbaum has suggested that the space average color of the display is the key image statistic that governs the system's color correction. Whittle and Challands (1969; Walraven, 1976) have suggested that the signal at the edges of the components of the image is the key image statistic that governs the system's color correction. We plan to use measurements of observer's ability to classify images to test a third color constancy algorithm. Maloney and Wandell (1986) suggested that color constancy can be obtained by using the spatial correlation of receptor responses.